Hugging Face's logo

zhiyang1
/
SZ_multinode_vlm_dc-vae-f32c32-768_group-7_no-self-attn-lora_residual_attn

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
d78e561
·
verified ·
1 Parent(s): da0cafb

Upload folder using huggingface_hub

Browse files
Files changed (46) hide show
  1. checkpoint-55000/config.json +79 -0
  2. checkpoint-55000/generation_config.json +6 -0
  3. checkpoint-55000/model.safetensors +3 -0
  4. checkpoint-55000/optimizer.pt +3 -0
  5. checkpoint-55000/rng_state_0.pth +3 -0
  6. checkpoint-55000/rng_state_1.pth +3 -0
  7. checkpoint-55000/rng_state_10.pth +3 -0
  8. checkpoint-55000/rng_state_11.pth +3 -0
  9. checkpoint-55000/rng_state_12.pth +3 -0
  10. checkpoint-55000/rng_state_13.pth +3 -0
  11. checkpoint-55000/rng_state_14.pth +3 -0
  12. checkpoint-55000/rng_state_15.pth +3 -0
  13. checkpoint-55000/rng_state_2.pth +3 -0
  14. checkpoint-55000/rng_state_3.pth +3 -0
  15. checkpoint-55000/rng_state_4.pth +3 -0
  16. checkpoint-55000/rng_state_5.pth +3 -0
  17. checkpoint-55000/rng_state_6.pth +3 -0
  18. checkpoint-55000/rng_state_7.pth +3 -0
  19. checkpoint-55000/rng_state_8.pth +3 -0
  20. checkpoint-55000/rng_state_9.pth +3 -0
  21. checkpoint-55000/scheduler.pt +3 -0
  22. checkpoint-55000/trainer_state.json +0 -0
  23. checkpoint-55000/training_args.bin +3 -0
  24. checkpoint-56000/config.json +79 -0
  25. checkpoint-56000/generation_config.json +6 -0
  26. checkpoint-56000/model.safetensors +3 -0
  27. checkpoint-56000/optimizer.pt +3 -0
  28. checkpoint-56000/rng_state_0.pth +3 -0
  29. checkpoint-56000/rng_state_1.pth +3 -0
  30. checkpoint-56000/rng_state_10.pth +3 -0
  31. checkpoint-56000/rng_state_11.pth +3 -0
  32. checkpoint-56000/rng_state_12.pth +3 -0
  33. checkpoint-56000/rng_state_13.pth +3 -0
  34. checkpoint-56000/rng_state_14.pth +3 -0
  35. checkpoint-56000/rng_state_15.pth +3 -0
  36. checkpoint-56000/rng_state_2.pth +3 -0
  37. checkpoint-56000/rng_state_3.pth +3 -0
  38. checkpoint-56000/rng_state_4.pth +3 -0
  39. checkpoint-56000/rng_state_5.pth +3 -0
  40. checkpoint-56000/rng_state_6.pth +3 -0
  41. checkpoint-56000/rng_state_7.pth +3 -0
  42. checkpoint-56000/rng_state_8.pth +3 -0
  43. checkpoint-56000/rng_state_9.pth +3 -0
  44. checkpoint-56000/scheduler.pt +3 -0
  45. checkpoint-56000/trainer_state.json +0 -0
  46. checkpoint-56000/training_args.bin +3 -0
checkpoint-55000/config.json ADDED
@@ -0,0 +1,79 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ar_steps": 1,
3
+ "architectures": [
4
+ "DiffVLMDiffusion"
5
+ ],
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 151643,
8
+ "condition_layer": -1,
9
+ "eos_token_id": 151645,
10
+ "hidden_act": "silu",
11
+ "hidden_size": 1536,
12
+ "image_token_id": 151655,
13
+ "img_cross_attention_dim": 2048,
14
+ "img_diffuser_depth": 2,
15
+ "img_ffn_dim_multiplier": null,
16
+ "img_hidden_size": 1536,
17
+ "img_multiple_of": 256,
18
+ "img_norm_eps": 1e-05,
19
+ "img_num_attention_heads": 12,
20
+ "img_num_kv_heads": 12,
21
+ "img_qk_norm": true,
22
+ "in_channels": 32,
23
+ "initializer_range": 0.02,
24
+ "inject_img_diffuser": false,
25
+ "input_size": 32,
26
+ "intermediate_size": 8960,
27
+ "layer_group_size": 7,
28
+ "layerwise_start_idx": 0,
29
+ "lora_alpha": 128,
30
+ "lora_bias": "none",
31
+ "lora_dropout": 0.05,
32
+ "lora_enable": false,
33
+ "lora_r": 64,
34
+ "max_position_embeddings": 32768,
35
+ "max_window_layers": 28,
36
+ "model_type": "qwen2_vl",
37
+ "non_linearity": 1,
38
+ "norm_elementwise_affine": true,
39
+ "num_attention_heads": 12,
40
+ "num_hidden_layers": 28,
41
+ "num_key_value_heads": 2,
42
+ "patch_size": 2,
43
+ "repa_coeff": 0.1,
44
+ "repa_layers": "2",
45
+ "repa_shared": false,
46
+ "rms_norm_eps": 1e-06,
47
+ "rope_scaling": {
48
+ "mrope_section": [
49
+ 16,
50
+ 24,
51
+ 24
52
+ ],
53
+ "rope_type": "default",
54
+ "type": "default"
55
+ },
56
+ "rope_theta": 1000000.0,
57
+ "sample_size": 128,
58
+ "sampling_steps": 28,
59
+ "sliding_window": null,
60
+ "tie_word_embeddings": true,
61
+ "torch_dtype": "bfloat16",
62
+ "transformers_version": "4.47.0",
63
+ "use_cache": true,
64
+ "use_repa": false,
65
+ "use_residual_attn": true,
66
+ "use_sliding_window": false,
67
+ "vae_path": "mit-han-lab/dc-ae-f32c32-sana-1.1-diffusers",
68
+ "video_token_id": 151656,
69
+ "vision_config": {
70
+ "hidden_size": 1536,
71
+ "in_chans": 3,
72
+ "model_type": "qwen2_vl",
73
+ "spatial_patch_size": 14
74
+ },
75
+ "vision_end_token_id": 151653,
76
+ "vision_start_token_id": 151652,
77
+ "vision_token_id": 151654,
78
+ "vocab_size": 151936
79
+ }
checkpoint-55000/generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 151643,
4
+ "eos_token_id": 151645,
5
+ "transformers_version": "4.47.0"
6
+ }
checkpoint-55000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a39653bd8d2459d594f366bb7c87deb15d196a323032472ce5fafabd10279d17
3
+ size 4539487822
checkpoint-55000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8319f7da646698f6f60e8de68feacfa74206659f12140c24548720208e7ca5f9
3
+ size 6632345482
checkpoint-55000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:34faf9c248c6a11b29ad92271f184eb0ac1e4c3bc45eacbd34129b80adf3d26c
3
+ size 15984
checkpoint-55000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b9c47ecf2b1d5b22ab8995f59e45151ff169ce14ce3cacfff21e7320a78e8f78
3
+ size 15984
checkpoint-55000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4f16ab07ce73564e590e6d3de86456e846d546be9c13c7ffd6148c8014185595
3
+ size 15997
checkpoint-55000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:df5a6727d316916fad7d6c2e7636e545783740c88eea5e98669845aca4965edb
3
+ size 15997
checkpoint-55000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:117a7021059d3993aeff715267a7a5e55b29a04171f071f979820451689e1a04
3
+ size 15997
checkpoint-55000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d7b92195b700daa9df85e614ddde881e8ffde7e7dffe58de1ed56a22aa55900f
3
+ size 15997
checkpoint-55000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:59bdc1adf76c95d09cd89c18036b51297f3826f718fcd18e0e691409cce4d9bc
3
+ size 15997
checkpoint-55000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c14d45f9552b2e6740534e191d9840d36a9bf5335dd95e209d7150d67f578f67
3
+ size 15997
checkpoint-55000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a0aa2b5feaaa103e748731429713ff73ee7488aa3972ea52dd29bfd7e5267c59
3
+ size 15984
checkpoint-55000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1005ba4d49f813f8c30fcac9e93240140288560a23befe30bbf0f9c65722004d
3
+ size 15984
checkpoint-55000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:754d501a1e5bb1b20875870997e472694d895b003d5d93e047ec6ceba87b70b8
3
+ size 15984
checkpoint-55000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe13cac7011dd8a05a72ff479540c9f082658d4b05f9cc916b841c58279cd441
3
+ size 15984
checkpoint-55000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:793191be436e1a312b8ebc8f89e514a5b3a24a4047e9709cee0b3a6150b163f2
3
+ size 15984
checkpoint-55000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a2a55c3d538a951198055714649367b17b134c224d89ab9dc2813970e63e830f
3
+ size 15984
checkpoint-55000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:20f929aaaf852d726d3b8e1a8936302ce6341e3dfbf0dcfb75b40e6818dcb46d
3
+ size 15984
checkpoint-55000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:27c84721bc71cdaff6f816f8b42bc89a5061017a68eaec23a52b57efaa92d21e
3
+ size 15984
checkpoint-55000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5fffc112bc488ea81b7e648b7dd97f0527bdf7c1596fd9caad9a9e3211d3fd41
3
+ size 1064
checkpoint-55000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-55000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:244bcdc3c5e05f9baca2a3010b6b7e5aca1e984909e26bd390ccb0bde66a432f
3
+ size 6008
checkpoint-56000/config.json ADDED
@@ -0,0 +1,79 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ar_steps": 1,
3
+ "architectures": [
4
+ "DiffVLMDiffusion"
5
+ ],
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 151643,
8
+ "condition_layer": -1,
9
+ "eos_token_id": 151645,
10
+ "hidden_act": "silu",
11
+ "hidden_size": 1536,
12
+ "image_token_id": 151655,
13
+ "img_cross_attention_dim": 2048,
14
+ "img_diffuser_depth": 2,
15
+ "img_ffn_dim_multiplier": null,
16
+ "img_hidden_size": 1536,
17
+ "img_multiple_of": 256,
18
+ "img_norm_eps": 1e-05,
19
+ "img_num_attention_heads": 12,
20
+ "img_num_kv_heads": 12,
21
+ "img_qk_norm": true,
22
+ "in_channels": 32,
23
+ "initializer_range": 0.02,
24
+ "inject_img_diffuser": false,
25
+ "input_size": 32,
26
+ "intermediate_size": 8960,
27
+ "layer_group_size": 7,
28
+ "layerwise_start_idx": 0,
29
+ "lora_alpha": 128,
30
+ "lora_bias": "none",
31
+ "lora_dropout": 0.05,
32
+ "lora_enable": false,
33
+ "lora_r": 64,
34
+ "max_position_embeddings": 32768,
35
+ "max_window_layers": 28,
36
+ "model_type": "qwen2_vl",
37
+ "non_linearity": 1,
38
+ "norm_elementwise_affine": true,
39
+ "num_attention_heads": 12,
40
+ "num_hidden_layers": 28,
41
+ "num_key_value_heads": 2,
42
+ "patch_size": 2,
43
+ "repa_coeff": 0.1,
44
+ "repa_layers": "2",
45
+ "repa_shared": false,
46
+ "rms_norm_eps": 1e-06,
47
+ "rope_scaling": {
48
+ "mrope_section": [
49
+ 16,
50
+ 24,
51
+ 24
52
+ ],
53
+ "rope_type": "default",
54
+ "type": "default"
55
+ },
56
+ "rope_theta": 1000000.0,
57
+ "sample_size": 128,
58
+ "sampling_steps": 28,
59
+ "sliding_window": null,
60
+ "tie_word_embeddings": true,
61
+ "torch_dtype": "bfloat16",
62
+ "transformers_version": "4.47.0",
63
+ "use_cache": true,
64
+ "use_repa": false,
65
+ "use_residual_attn": true,
66
+ "use_sliding_window": false,
67
+ "vae_path": "mit-han-lab/dc-ae-f32c32-sana-1.1-diffusers",
68
+ "video_token_id": 151656,
69
+ "vision_config": {
70
+ "hidden_size": 1536,
71
+ "in_chans": 3,
72
+ "model_type": "qwen2_vl",
73
+ "spatial_patch_size": 14
74
+ },
75
+ "vision_end_token_id": 151653,
76
+ "vision_start_token_id": 151652,
77
+ "vision_token_id": 151654,
78
+ "vocab_size": 151936
79
+ }
checkpoint-56000/generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 151643,
4
+ "eos_token_id": 151645,
5
+ "transformers_version": "4.47.0"
6
+ }
checkpoint-56000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c1ddcf4525068e1aadab348f213c6d766e25d7bedb6e3de007f0dbe363cca866
3
+ size 4539487822
checkpoint-56000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aafa2790e46273c8174289fb7187ddd6e215898e8d840154276ca0d4e6d2b537
3
+ size 6632345482
checkpoint-56000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:41556ef98234aa8e08114627e593827920feadf21c9ddac33dd423295091066b
3
+ size 15984
checkpoint-56000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a10ccf2527fded6fc49476ce90321170dc58b52c90dead15de48ef9b20789351
3
+ size 15984
checkpoint-56000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8b2055f8f7ff4bb18e48d862a63213b5f8fd858a35434b1fc43d225d85714ea3
3
+ size 15997
checkpoint-56000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a41e501b5b4702073321fc76be55334876f6509116ea1d17ba1b17f1a06e4b1
3
+ size 15997
checkpoint-56000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e58aee8645ebeb102636fe39b8a097040d1e615da4264a7ed0c7aec2fb1246dd
3
+ size 15997
checkpoint-56000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1095fbac54d5f48d046564910546acd4656402cdb706164eaa87e9b8c4ba3905
3
+ size 15997
checkpoint-56000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e86e25be1ffe21dbaad7cc81e148722f0530e7012c64200432e3ce195872f0cd
3
+ size 15997
checkpoint-56000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6351751abb43f8cd4a5da409505258a30896b49e8e3897a2ef5e334dcbcc01e1
3
+ size 15997
checkpoint-56000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f03a40358ab668f65a41b62642069a8af127f586b92670a7113ff23972d32854
3
+ size 15984
checkpoint-56000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fb37b34a9edc173382ca25409f870e20dd71ebe42e4358c0006f2e994a0047ca
3
+ size 15984
checkpoint-56000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4440009593f0cff06507ec9bb81c474aa5161c44e10cba4d66ff52f9532eada5
3
+ size 15984
checkpoint-56000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:178d65b168a8633f9cfdf907d12db09717b2db740d676dbd9f88ba571d159004
3
+ size 15984
checkpoint-56000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:be5247011f37e0dd8e266e175f34d429837c3cc366406e95c6629226a1888647
3
+ size 15984
checkpoint-56000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:204747fbd76e937a15360d9c61af167f7463765b528d49c9d6c8dde61a1c394e
3
+ size 15984
checkpoint-56000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2cb63e200a70f0b1d3af047c60f3c6f0584f82dfcc33507885076ab90a0616ee
3
+ size 15984
checkpoint-56000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:34feefab04768e338d4fa32bd7d61647562ed65cf8cf1cca8f7595ec663e98d4
3
+ size 15984
checkpoint-56000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ba4f33232eebcc1e79f822b2dde4537e011cb3134ce17d90d2885a430e71f68c
3
+ size 1064
checkpoint-56000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-56000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:60ea994133b53f46c701343e28bbe591bd274de0589899746f4d77571d80907b
3
+ size 6008